DESCRIPTION Schizophrenia patients have difficulty interpreting facial emotions. This proposal examines their ability to implicitly learn facial stimuli. It uses a paradigm developed by Lewicki and his colleagues in which people learn non-salient covariations, or relations between features that are presented together, without having express knowledge of these co-variations. First, subjects will implicitly learn a co-variation between word-emotion pairs and their spatial location on a video screen. Because prior work shows normal implicit learning of verbal material in schizophrenia it is expected that patients will perform normally on this task. Second, pictures of faces replace verbal material. Thus, subjects will associate a covariation between a face- emotion pair and its spatial location on a video screen. Because this task does not require patients to process internal facial features, it is expected that patients will perform normally on this task as well. Finally, subjects will be presented with a covariation between internal facial features and emotion. Because learning the covariation requires perception of internal facial features, it is expected that the patients will be impaired. If patients do not show normal learning, it will be the first demonstration of abnormal implicit learning of faces in this group. If patients perform normally on this implicit learning tasks, then this finding demonstrates an intact system for learning covariations. Further, this method of training would be a potential therapeutic means for teaching schizophrenia patients how to interpret social cues.